Hugepage COW

[David Gibson]

On Mon, Mar 21, 2005 at 10:26:39AM -0600, Adam Litke wrote:
> On Wed, 2005-03-16 at 21:48, David Gibson wrote:
> > As far as the consolidation patch goes, lack of testing was the main
> > objective reason for holding back.  So if you could test on x86_64 and
> > ia64 too, that would be great.  wli had some objections to the patch
> > when I first posted which I didn't and don't really understand, and
> > from conversations with akpm, I'm certainly thinking of just
> > re-sending it.
> 
> I am in the process of obtaining an x86_64 box to test these on.  IA64
> hardware could prove more difficult to find though.

That's fine.  Testing on three archs is still better than two, even if
four would be better yet :)

> > COW will be a bit more of a political shitfight, I suspect.  I'd like
> > to at least hold off until the consolidation is merged, which makes
> > the COW much easier.  We'll also need to implement the necessary
> > arch-hooks for COW on every platform.  Speaking of which, did you
> > implement the i386 hooks?  I thought I only did COW for ppc64, so far,
> > although on top of the consolidation patch the amount of arch code is
> > vastly reduced.
> 
> The version of your cow that I have has no hooks (not even for ppc64). 
> I applies on top of the consolidate patch.  It is working fine on x86
> but I am investigating a possible problem on ppc64.

Ah, yes, just checked and I see now.  It only needs arch hooks for
those archs which define ARCH_HAS_SETCLEAR_HUGE_PTE (in the
consolidation patch).  Which I think is just sparc64, sh and sh64.

-- 
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/people/dgibson